 it's pretty good. So, as mentioned, I'm Colin Charles from the Iberia DB Project. I'm one of the lucky few people to get paid to work on this database. I always like to start with the Alexa rankings. They have 500 top sites in the Iraqi globally. I picked the 20 that you can fit up on the slide. Out of 20, 19 of them run some variant of my SQL. Nobody really runs my SQL in production without patches. Facebook, number two, basically started the web scale SQL project with Twitter number eight as well as LinkedIn number 12 and what they do is they take my SQL 56 and add on hundreds of patches to it and they use that as a base to build their own distributions of my SQL. Companies like Google, YouTube, AutoGoogle, Wikipedia, they all run MariaDB, Wikipedia soft MariaDB, Google, obviously a very customized version of MariaDB and lately I'm sure some of you may have heard of Amazon's Aurora RDS service. Anybody heard of that? Use AWS. We're working quite actively with Amazon and Aurora as well. So on the top 20 websites out there and some of the sites that don't make it include Tumblr, WordPress, .com and so forth, they all run somewhere into my SQL or most likely even MariaDB nowadays. We are a fully GPL v2 branch of my SQL. Some would call us a fork but every month or every second month an article makes a release, we actually make a merge. So you get the features that we have plus the features that Oracle has fixed with my SQL and a lot of the stuff we do is focused on community development and I think that's really important for us. Communities are key to us. We are on launchpad, we are on GitHub. It is as easy as branching the code base, parking it and then saying you'd like a call request submitted and you don't even need to sign a contributor agreement today. We accept BSD licensed code. This is one of the main reasons why it's hard for WebScale SQL features to make it back up into upstream my SQL. It's because you have to sign a contributor agreement. The contributor agreement is something that many corporate lawyers don't like people signing. We obviously feature enhanced. We spend a long time working on MariaDB. We definitely application compatible. So today you basically run my SQL, you run MariaDB, you release a drop in replacement, your application will no no different and we are definitely feature complete. We actually celebrated our fifth birthday on February 1st. We've been around now for five years shipping GA releases. In that time we've had over 2 million downloads and our average is about 200,000 downloads from the website a month. This is not counting the downloads you get from the distributions and BSDs and so forth. In that short span of time we've made over 88 releases of software in the server. 61 GA releases, 7 RCs, 11 patas, 9 alphas. Just in the past week we made two more GA releases. So we're pretty active for a very relatively small team compared to the engineering prowess that Oracle could hire for my SQL. We've made many favorites. In the last five years the MySQL world has seen MySQL 55, MySQL 56 and the top of MySQL 57. In the past five years we've done 51, 52, 53, 55, 10, 0, a couple of GA releases and we're on track to release 10-1. Sometimes this year it's already in beta. So final releases in quite a short period of time. Why MariaDB? Why LibreOffice? A lot of this has to do with vendor independence and freedom from vendor locking is truly important. MySQL has a single owner, a single vendor. MariaDB has a foundation to ensure that this sort of thing never happens again. It also has a cooperation backing it. So you can think of this very much like the Mozilla cooperation and foundation to that extent. There will never be another fork on MariaDB thanks to the foundation and it's relatively strong considering that there are many other people supporting MariaDB outside of the cooperation. I would say that the ecosystem development for MySQL is that it's most vibrant now. All thanks to the owner of MySQL. Many people are working on branches and our forks. The community gets any feature you want, you think is useful, you can be getting it shipping in a major release. That's how we've gotten contributions for things like multi-sales replication and so forth. And if you're a storage engine vendor like TopiTag making new strides with indexes like factory indexes instead of regular B plus three indexes or you think you want to make a no SQL storage engine against say a Cassandra cluster or a Padoon cluster, we provide the ability for you to do that relatively easily and we ship all these engines that allow connectivity to other places. We'd like to be in the middle that allows you to connect to SQL, no SQL, big data storage layers. Pretty much all the early MySQL developers joined MariaDB in the early days including obviously MySQL's creator Michael Monti Vidinius. We have many many contributors and the ratio is actually pretty impressive. You can go to Launchpad and check but you to get common right access you need to be what is referred to as a Maria captain. We recently been migrating to GitHub so there are a bunch of people moving here and there and you'd realize that we have along the lines of about 25 external contributors, so people in the community, 18 who work for the corporation and five who work for the foundation. So the split is actually pretty even and actually leading towards more community contributors who are working at major web companies and so forth actually pushing code to MariaDB. Besides the server, MariaDB also focuses on LGPL client libraries or connectors as you may. One of the major ones we actually had to obviously make was DC connector so you could build things on top of it. Why LGPL and not GPL? Because when you go to data of mysql.com and download a connector you're getting a GPL client with false exceptions which allow you to use the connector for false applications but if you're going to embed it you would actually have to pay a license fee to get past GPL as part of your license. We give you LGPL connectors so you never have to pay a license fee when you want to embed be it for C, be it for Java or be it for OUBC and there are community connectors that are being built and we will eventually when they get to a virtual level bring them back into the project maybe in the C++ area and curl. C++ mainly for the library office project because they also want a database connector for the base application. We also really like to focus on making the things you can get in mysql enterprise open. And a great example of why we do things in the open. We initially saw a thread pool in mysql enterprise. A thread pool is basically allowing you to open up a limited pool of threads that can be reused because the mysql model is to open up one thread per connection and then this is bad for CPU locality. We made a thread pool, we released this open source. Another company, Percona, improved the thread pool and because they had a customer that needed some features and then we managed to take their features back again. This is something that I'm pretty sure many of you in the open source will understand. Having Percona out there in the open makes a lot of sense. Many eyes also make shallow bugs. So when there are enterprise features that you have to pay for, you don't really know if you're going to hit it in a bargain. But with the open version, at least you know that you're getting many, many people using and all working on it. Our knowledge base is also fully open content and also license friendly, GMDL as well as credit comments. You may want to check it out. But MariaDB has grown beyond just being a server and connectors. Recently, we've also come up with a level seven proxy that actually basically understands the mysql protocol. It allows you to route connections or statements. It's a pluggable router and the best part is is you can also write your plugins relatively easily you would see and this product really only became GA in January and there's already another company that's started up and it's March now. So that's literally two months and another company started taking use of this and putting it, putting dashboards and so forth out there for you to call Maxfair. Then we have people who want to use binary log relays and you know the database will always have binary logs long before the system B will happen. And we like binary logs because we can replay binary logs on another machine. We can literally grab through them with a tool. So we really do like when you have multiple states, you can use Maxfair as a binary log relay server to actually push to multiple states. The possibilities with this are generally kind of endless and the fact that you can also load balance to galore clusters so you don't have to use any proxies kind of handy. A very common use case for this is to allow the proxy not only to write to MariaDB or mysql backend but also another backend say MongoDB because you can actually write the queries rewrite on the fly and actually insert it into another database and they are plugins already available for this. You can get MariaDB at MariaDB.org and presumably all of you run Linux on your servers and you'll find that it's a default in things like Rails 7, Santos, Sousa and Price 12, OpenSousa, etc. In fact, Fedora was one of the first distributions to start shipping MariaDB 5.5 and recent releases of Fedora actually had that upgrade go from Fedora MariaDB 5.5 to MariaDB 10. And the pod count, we were actually really worried that users would complain that there would be huge amounts of bugs and so forth but actually the pod count didn't increase. People are using MariaDB 10 as a job and replacement without having any issues and these include apps that are built against it as well. So default might bring mysql in pretty much many of the distributions. When you ask for mysql, you get MariaDB for free. So you may not know that you're already using MariaDB to never start a client. We also really love participating in Google summer of code and from the first year we participated, we got a committer. The second year we managed, someone within the foundation managed a higher person who came out of it and we just got accepted for 2015 as well and all this is students writing code and our shipping and parent releases of MariaDB. Now you might be wondering who is using MariaDB. Most likely you are using it even without knowing you didn't put up your hands on it because pretty much all the next distributions will give it to you for free as mysql. You start it up, type in mysql as well. Google was migrating to MariaDB last year but I'm very happy to say that throughout the most of last year they've actually pretty much completed the migration to MariaDB 10. They funded the parallel application work we do and parallel application is actually pretty handy because you can have parallel apply to your slave servers or parallel apply even when you run. They also do code reviews, generally they all happen on public, they do QA, they do testing and you know we have to thank Edward Snowden for telling us that the NSA is hacking lots of servers so Google has decided that they want end to end encryption for everything which is why they want to search, they want HTTPS and internally they also like to run things encrypted and they actually wrote a feature called table space encryption that allows you to encrypt your entire inner DB table space and then we obviously have key management externally and they contributed this code for MariaDB 10.1 so not only are they consuming MariaDB, they're also extending MariaDB and making it better for you and for the first time in the MySQL world and MySQL will be 20 years old in May we have actually done encryption for table spaces. This should be a huge huge thing for enterprise use cases or anybody who cares about their privacy and early testing suggests performance degradation of no more than 10% on regular hardware, not even modern hardware just regular they went from MySQL 5.1 to MariaDB 5.5 and also started migrating pretty much to MariaDB 10 in recent times. The initial reasoning was the better optimizer and that's the most true, we also shipped extra DB though the extra DB in the DB delta is getting a lot smaller but we can be there, there are major improvements for query times, they were getting anytime between 4 to 15 percent query time improvements which would help things load faster and why would you care if things load faster? I mean in Singapore you have one gigabit internet to your home right? But if you think about the rest of the world where there are things like Wikipedia 0 and Facebook 0 where they actually basically tell carriers not to charge on mobile devices, having a Wikipedia page load 15 percent pasta means the carrier pays less, a user in some free thermal country doesn't wait 15 percent longer and so forth, this is actually pretty impactful for real world use cases. Anybody here use Kakaotalk for messaging or is everybody a WhatsApp user? So they generally are very famous if you're in the Korean or K-pop scene of space, if you like, the Hollywood, they have a lot of active users, they started with MariaDB 5.5, they also moved to MariaDB 10 and keep in mind MariaDB 10 only became GA last April so all these all these migrations that happened within the span of less than and not only are they actively using the server, they're also doing code contributions so they've actually backbotted in a DVD fragmentation, they backbotted these from web scale and so forth so we not only have happy users, we also have happy developers. One of the interesting migrations that I've seen lately is Greeds, they didn't just migrate from my SQL to MariaDB but they actually ditched Oracle Rack which is really expensive for MariaDB Galarock cluster and they claimed that they were getting a lot better lower the GCO as well as admin costs and you don't need shared storage any longer. These people are in the e-commerce greeting card space but we see similar things from people in the web hosting space, the telco space, people just love migrating to MariaDB so you know Alcatel Lucent 10s, mix user connectors and so forth. A lot of talk is about 5.7. 5.7 is not here today, always remember that. 5.7 has been talked about since last year, it'll be talked about again this year and it may show its hit end of the CO next year. We're giving you innovation today that you're going to see in 5.7 tomorrow, we've had a lot of these features already in 10.0. Multisource Replication is a lab's release that you can get from labs.mysql.com and we've been shipping it for a long time, it was contributed by Tava, it allows you multiple masters to write to a single slave. You can either use this for analytics like front class games where it has nodes and AWS Singapore does, you can use this for complete backups, you can use this for whatever you really want. We've been shipping GIS functionality so you can actually now make use of latitudes and longitudes and transformations and get parts between them since Marinev5.3. We've shipped NoSQL APIs like Handler Socket that skip the entire SQL layer that write directly to InnoDB for create read, update, delete operations and we also provide dynamic columns which allow each and every row in your table to have different amounts of columns, very NoSQL like so to speak and that feature also allows you to get the output, you can query the output and get it back as JSON output so you can use that in your web app relatively easily. Already mentioned power replication, we've done GTID better. Web scale SQL had to do GTID via adding a little table, the Maya scale we have there in GTID which is global transaction IDs is that you shut down your entire topology and restart it. We just append the GTID to the replication packet so you can actually upgrade in production in C2. And we also shipped the correct engine which can now read JSON and JSON and you can try this today. Not only does it read JSON the standard, it also reads the MongoDB Veson so you can actually read MongoDB Veson files with ease today. Where do we go from here? Today if you're playing around with the open stack worlds or any of these cloud worlds, they will generally recommend that you want to have something like Galera Cluster. Galera Cluster supports fully synchronous replication whereas Maya SQL and MariaDB only support asynchronous replication or semi-synchronous Maya plugin. Fully synchronous means that all nodes get the comment or no nodes get the comment. And we've made life easier. In 10.1 we've integrated Galera Cluster into MariaDB. So now there's only one download. Not only is there only one download, you can turn on and off the feature as and when you feel like it, which is a huge bonus. So now if you have multi-geographical centered environments, you may want to have synchronous replication, say in Singapore asynchronous to Tokyo, maybe fully synchronous again to the US. All this is not fully possible with MariaDB 10.1. Not only can we encrypt table spaces thanks to Google, we also took in a contribution from another company called Ebrida allows you to encrypt just individual tables. So don't encrypt everything, just encrypt some tables. Now some of you may think this is not the best idea. Encrypting is tables because it leaves all the control into the hands of an admin, but in many shops the admin really does that control anyway, especially if you're small. We also invest heavily in instrumentation around the information schema, which is just a metadata engine as opposed to performance schema, which actually has performance heat. We really are focusing security. We've put in the past validation plugins. We can now check your password against Pam Cracklet to tell you your passwords are too weak. Also giving you cover authentication in addition to the Pam authentication that we've already provided and the LDAP authentication. And we also integrated pretty much every web scale patch that makes sense to the greater MariaDB environment. So generally speaking, today you can download MariaDB 10.1 Beta. In five years we've come pretty far for a relatively small team of developers and it's largely thanks to the fact that there are a community of users, not only users out there playing around with the database for fun, but also users in production in large companies believing that we can do MariaDB. So five years, still a branch, not a fork. Check us out if you haven't already. With that I say thank you and I guess I'm open to questions. We're a little tight on time, so we have one focus question. So technically, could you give a quick comparison between MySQL Enterprise cluster and Galera cluster? So you want a quick, I need to clarify this quick difference between MariaDB Enterprise cluster and Galera cluster. So MariaDB MySQL Enterprise cluster or MySQL cluster carrier grade is something called NDB cluster. That's a technical name in the storage engine. That is a different storage engine compared to NDB. NDB cluster requires you to have literally a minimum of five nodes. It does not need to run in NDB and when you want to do things like complex joins, it still happens to go over the network because of the data model. Galera cluster does the joins locally and then streams that it's going to write in the streams that are after the commit. So NDB cluster is used in the telcospace quite a bit and pretty much only in the telcospace. Galera cluster can be used in the telcospace and every other space and the main player to claim and benefit I would say is the fact that it actually runs in NDB. Now nothing would be complete because there's always trade-offs. With NDB cluster, you could probably grow to N amount of nodes and N is the number that is probably large. With Galera cluster, because of synchronous replication, if you keep on adding nodes, say you hit 15 nodes and N above, things generally stop slowing down because it has to do the writes and every nodes. The reads and the reads and the writes will definitely slow down. That's when you start using things like max scale, orange and proxy because you send writes to just a few nodes and reads to all the other nodes so you end up using low balance. So that's a general thing for the virus. I may do this much more, but now is not the time. Thank you very much. Quick update on the schedule.